Research map / Architectures and efficiency / RNNs and normalization
House of Cans: Covert Transmission of Internal Datasets via Capacity-Aware Neuron Steganography
Xudong Pan, Shengyao Zhang, Mi Zhang, Yifan Yan, Min Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase stolen samples from 512 to 4096 | Cans · SpeechCommand | MSE: 2.7 × 10−4 -> 4.8 × 10−4 | Hurt |
| increase stolen samples from 512 to 4096 | Cans · SpeechCommand | carrier Delta Perf: −0.43% -> −0.35% | No clear effect |
| steal three datasets simultaneously instead of Speech alone | Cans · SpeechCommand 1024 samples | MSE: 3.5 × 10−4 -> 4.1 × 10−4 | No clear effect |
| increase FaceScrub stolen images from 1024 to 32768 | Cans · FaceScrub | SSIM: 1.0 -> 0.6 | Hurt |
| weight pruning ratio beta 0.3 on carrier model | Cans · FaceScrub 512 images | SSIM of decoded images: 0.994 -> 0.8 | Hurt |
| weight pruning ratio beta 0.3 on carrier model | Cans · FaceScrub 512 images | carrier performance: over 10% | Hurt |
| remove fusion (weight pool restoration) from pruned model decoding (ablation) | Cans · FaceScrub pruning | secret model reconstruction | Hurt |
| use Cans instead of Sign/Covariance/LSB baselines at N=64+ | Cans · FaceScrub | hiding feasibility: Nmax = 76 | Improved |
| use LSB encoding at N=64 instead of Cans | LSB baseline · FaceScrub | carrier accuracy | Hurt |
| use Cans instead of Sign and Covariance encoding | Cans · FaceScrub/CIFAR-10 | SSIM | Improved |
Closest papers by meaning
- Hiding Images in Diffusion Models by Editing Learned Score Functions (CVPR 2025 · similarity 0.83)
- Revisiting the Information Capacity of Neural Network Watermarks: Upper Bound Estimation and Beyond (AAAI 2024 · similarity 0.82)
- StegFormer: Rebuilding the Glory of Autoencoder-Based Steganography (AAAI 2024 · similarity 0.82)
- Delving into Differentially Private Transformer (ICML 2024 · similarity 0.82)
- Purified and Unified Steganographic Network (CVPR 2024 · similarity 0.82)
- Rethinking DP-SGD in Discrete Domain: Exploring Logistic Distribution in the Realm of signSGD (ICML 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
